| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * Copyright (c) 2023 Paul B Mahol | ||
| 3 | * | ||
| 4 | * This file is part of FFmpeg. | ||
| 5 | * | ||
| 6 | * FFmpeg is free software; you can redistribute it and/or | ||
| 7 | * modify it under the terms of the GNU Lesser General Public | ||
| 8 | * License as published by the Free Software Foundation; either | ||
| 9 | * version 2.1 of the License, or (at your option) any later version. | ||
| 10 | * | ||
| 11 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 14 | * Lesser General Public License for more details. | ||
| 15 | * | ||
| 16 | * You should have received a copy of the GNU Lesser General Public | ||
| 17 | * License along with FFmpeg; if not, write to the Free Software | ||
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 19 | */ | ||
| 20 | |||
| 21 | #include "libavutil/common.h" | ||
| 22 | #include "libavutil/float_dsp.h" | ||
| 23 | #include "libavutil/mem.h" | ||
| 24 | #include "libavutil/opt.h" | ||
| 25 | |||
| 26 | #include "audio.h" | ||
| 27 | #include "avfilter.h" | ||
| 28 | #include "formats.h" | ||
| 29 | #include "filters.h" | ||
| 30 | |||
| 31 | enum OutModes { | ||
| 32 | IN_MODE, | ||
| 33 | DESIRED_MODE, | ||
| 34 | OUT_MODE, | ||
| 35 | NOISE_MODE, | ||
| 36 | ERROR_MODE, | ||
| 37 | NB_OMODES | ||
| 38 | }; | ||
| 39 | |||
| 40 | typedef struct AudioRLSContext { | ||
| 41 | const AVClass *class; | ||
| 42 | |||
| 43 | int order; | ||
| 44 | float lambda; | ||
| 45 | float delta; | ||
| 46 | int output_mode; | ||
| 47 | int precision; | ||
| 48 | |||
| 49 | int kernel_size; | ||
| 50 | AVFrame *offset; | ||
| 51 | AVFrame *delay; | ||
| 52 | AVFrame *coeffs; | ||
| 53 | AVFrame *p, *dp; | ||
| 54 | AVFrame *gains; | ||
| 55 | AVFrame *u, *tmp; | ||
| 56 | |||
| 57 | AVFrame *frame[2]; | ||
| 58 | |||
| 59 | int (*filter_channels)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs); | ||
| 60 | |||
| 61 | AVFloatDSPContext *fdsp; | ||
| 62 | } AudioRLSContext; | ||
| 63 | |||
| 64 | #define OFFSET(x) offsetof(AudioRLSContext, x) | ||
| 65 | #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM | ||
| 66 | #define AT AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM | ||
| 67 | |||
| 68 | static const AVOption arls_options[] = { | ||
| 69 | { "order", "set the filter order", OFFSET(order), AV_OPT_TYPE_INT, {.i64=16}, 1, INT16_MAX, A }, | ||
| 70 | { "lambda", "set the filter lambda", OFFSET(lambda), AV_OPT_TYPE_FLOAT, {.dbl=1.f}, 0, 1, AT }, | ||
| 71 | { "delta", "set the filter delta", OFFSET(delta), AV_OPT_TYPE_FLOAT, {.dbl=2.f}, 0, INT16_MAX, A }, | ||
| 72 | { "out_mode", "set output mode", OFFSET(output_mode), AV_OPT_TYPE_INT, {.i64=OUT_MODE}, 0, NB_OMODES-1, AT, .unit = "mode" }, | ||
| 73 | { "i", "input", 0, AV_OPT_TYPE_CONST, {.i64=IN_MODE}, 0, 0, AT, .unit = "mode" }, | ||
| 74 | { "d", "desired", 0, AV_OPT_TYPE_CONST, {.i64=DESIRED_MODE}, 0, 0, AT, .unit = "mode" }, | ||
| 75 | { "o", "output", 0, AV_OPT_TYPE_CONST, {.i64=OUT_MODE}, 0, 0, AT, .unit = "mode" }, | ||
| 76 | { "n", "noise", 0, AV_OPT_TYPE_CONST, {.i64=NOISE_MODE}, 0, 0, AT, .unit = "mode" }, | ||
| 77 | { "e", "error", 0, AV_OPT_TYPE_CONST, {.i64=ERROR_MODE}, 0, 0, AT, .unit = "mode" }, | ||
| 78 | { "precision", "set processing precision", OFFSET(precision), AV_OPT_TYPE_INT, {.i64=0}, 0, 2, A, .unit = "precision" }, | ||
| 79 | { "auto", "set auto processing precision", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A, .unit = "precision" }, | ||
| 80 | { "float", "set single-floating point processing precision", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A, .unit = "precision" }, | ||
| 81 | { "double","set double-floating point processing precision", 0, AV_OPT_TYPE_CONST, {.i64=2}, 0, 0, A, .unit = "precision" }, | ||
| 82 | { NULL } | ||
| 83 | }; | ||
| 84 | |||
| 85 | AVFILTER_DEFINE_CLASS(arls); | ||
| 86 | |||
| 87 | ✗ | static int query_formats(const AVFilterContext *ctx, | |
| 88 | AVFilterFormatsConfig **cfg_in, | ||
| 89 | AVFilterFormatsConfig **cfg_out) | ||
| 90 | { | ||
| 91 | ✗ | const AudioRLSContext *s = ctx->priv; | |
| 92 | static const enum AVSampleFormat sample_fmts[3][3] = { | ||
| 93 | { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_DBLP, AV_SAMPLE_FMT_NONE }, | ||
| 94 | { AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE }, | ||
| 95 | { AV_SAMPLE_FMT_DBLP, AV_SAMPLE_FMT_NONE }, | ||
| 96 | }; | ||
| 97 | int ret; | ||
| 98 | |||
| 99 | ✗ | if ((ret = ff_set_common_formats_from_list2(ctx, cfg_in, cfg_out, | |
| 100 | ✗ | sample_fmts[s->precision])) < 0) | |
| 101 | ✗ | return ret; | |
| 102 | |||
| 103 | ✗ | return 0; | |
| 104 | } | ||
| 105 | |||
| 106 | ✗ | static int activate(AVFilterContext *ctx) | |
| 107 | { | ||
| 108 | ✗ | AudioRLSContext *s = ctx->priv; | |
| 109 | int i, ret, status; | ||
| 110 | int nb_samples; | ||
| 111 | int64_t pts; | ||
| 112 | |||
| 113 | ✗ | FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx); | |
| 114 | |||
| 115 | ✗ | nb_samples = FFMIN(ff_inlink_queued_samples(ctx->inputs[0]), | |
| 116 | ff_inlink_queued_samples(ctx->inputs[1])); | ||
| 117 | ✗ | for (i = 0; i < ctx->nb_inputs && nb_samples > 0; i++) { | |
| 118 | ✗ | if (s->frame[i]) | |
| 119 | ✗ | continue; | |
| 120 | |||
| 121 | ✗ | if (ff_inlink_check_available_samples(ctx->inputs[i], nb_samples) > 0) { | |
| 122 | ✗ | ret = ff_inlink_consume_samples(ctx->inputs[i], nb_samples, nb_samples, &s->frame[i]); | |
| 123 | ✗ | if (ret < 0) | |
| 124 | ✗ | return ret; | |
| 125 | } | ||
| 126 | } | ||
| 127 | |||
| 128 | ✗ | if (s->frame[0] && s->frame[1]) { | |
| 129 | AVFrame *out; | ||
| 130 | |||
| 131 | ✗ | out = ff_get_audio_buffer(ctx->outputs[0], s->frame[0]->nb_samples); | |
| 132 | ✗ | if (!out) { | |
| 133 | ✗ | av_frame_free(&s->frame[0]); | |
| 134 | ✗ | av_frame_free(&s->frame[1]); | |
| 135 | ✗ | return AVERROR(ENOMEM); | |
| 136 | } | ||
| 137 | |||
| 138 | ✗ | ff_filter_execute(ctx, s->filter_channels, out, NULL, | |
| 139 | ✗ | FFMIN(ctx->outputs[0]->ch_layout.nb_channels, ff_filter_get_nb_threads(ctx))); | |
| 140 | |||
| 141 | ✗ | out->pts = s->frame[0]->pts; | |
| 142 | ✗ | out->duration = s->frame[0]->duration; | |
| 143 | |||
| 144 | ✗ | av_frame_free(&s->frame[0]); | |
| 145 | ✗ | av_frame_free(&s->frame[1]); | |
| 146 | |||
| 147 | ✗ | ret = ff_filter_frame(ctx->outputs[0], out); | |
| 148 | ✗ | if (ret < 0) | |
| 149 | ✗ | return ret; | |
| 150 | } | ||
| 151 | |||
| 152 | ✗ | if (!nb_samples) { | |
| 153 | ✗ | for (i = 0; i < 2; i++) { | |
| 154 | ✗ | if (ff_inlink_acknowledge_status(ctx->inputs[i], &status, &pts)) { | |
| 155 | ✗ | ff_outlink_set_status(ctx->outputs[0], status, pts); | |
| 156 | ✗ | return 0; | |
| 157 | } | ||
| 158 | } | ||
| 159 | } | ||
| 160 | |||
| 161 | ✗ | if (ff_outlink_frame_wanted(ctx->outputs[0])) { | |
| 162 | ✗ | for (i = 0; i < 2; i++) { | |
| 163 | ✗ | if (s->frame[i] || ff_inlink_queued_samples(ctx->inputs[i]) > 0) | |
| 164 | ✗ | continue; | |
| 165 | ✗ | ff_inlink_request_frame(ctx->inputs[i]); | |
| 166 | ✗ | return 0; | |
| 167 | } | ||
| 168 | } | ||
| 169 | ✗ | return 0; | |
| 170 | } | ||
| 171 | |||
| 172 | #define DEPTH 32 | ||
| 173 | #include "arls_template.c" | ||
| 174 | |||
| 175 | #undef DEPTH | ||
| 176 | #define DEPTH 64 | ||
| 177 | #include "arls_template.c" | ||
| 178 | |||
| 179 | ✗ | static int config_output(AVFilterLink *outlink) | |
| 180 | { | ||
| 181 | ✗ | AVFilterContext *ctx = outlink->src; | |
| 182 | ✗ | AudioRLSContext *s = ctx->priv; | |
| 183 | |||
| 184 | ✗ | s->kernel_size = FFALIGN(s->order, 16); | |
| 185 | |||
| 186 | ✗ | if (!s->offset) | |
| 187 | ✗ | s->offset = ff_get_audio_buffer(outlink, 1); | |
| 188 | ✗ | if (!s->delay) | |
| 189 | ✗ | s->delay = ff_get_audio_buffer(outlink, 2 * s->kernel_size); | |
| 190 | ✗ | if (!s->coeffs) | |
| 191 | ✗ | s->coeffs = ff_get_audio_buffer(outlink, 2 * s->kernel_size); | |
| 192 | ✗ | if (!s->gains) | |
| 193 | ✗ | s->gains = ff_get_audio_buffer(outlink, s->kernel_size); | |
| 194 | ✗ | if (!s->p) | |
| 195 | ✗ | s->p = ff_get_audio_buffer(outlink, s->kernel_size * s->kernel_size); | |
| 196 | ✗ | if (!s->dp) | |
| 197 | ✗ | s->dp = ff_get_audio_buffer(outlink, s->kernel_size * s->kernel_size); | |
| 198 | ✗ | if (!s->u) | |
| 199 | ✗ | s->u = ff_get_audio_buffer(outlink, s->kernel_size); | |
| 200 | ✗ | if (!s->tmp) | |
| 201 | ✗ | s->tmp = ff_get_audio_buffer(outlink, s->kernel_size); | |
| 202 | |||
| 203 | ✗ | if (!s->delay || !s->coeffs || !s->p || !s->dp || !s->gains || !s->offset || !s->u || !s->tmp) | |
| 204 | ✗ | return AVERROR(ENOMEM); | |
| 205 | |||
| 206 | ✗ | for (int ch = 0; ch < s->offset->ch_layout.nb_channels; ch++) { | |
| 207 | ✗ | int *dst = (int *)s->offset->extended_data[ch]; | |
| 208 | |||
| 209 | ✗ | for (int i = 0; i < s->kernel_size; i++) | |
| 210 | ✗ | dst[0] = s->kernel_size - 1; | |
| 211 | } | ||
| 212 | |||
| 213 | ✗ | switch (outlink->format) { | |
| 214 | ✗ | case AV_SAMPLE_FMT_DBLP: | |
| 215 | ✗ | for (int ch = 0; ch < s->p->ch_layout.nb_channels; ch++) { | |
| 216 | ✗ | double *dst = (double *)s->p->extended_data[ch]; | |
| 217 | |||
| 218 | ✗ | for (int i = 0; i < s->kernel_size; i++) | |
| 219 | ✗ | dst[i * s->kernel_size + i] = s->delta; | |
| 220 | } | ||
| 221 | |||
| 222 | ✗ | s->filter_channels = filter_channels_double; | |
| 223 | ✗ | break; | |
| 224 | ✗ | case AV_SAMPLE_FMT_FLTP: | |
| 225 | ✗ | for (int ch = 0; ch < s->p->ch_layout.nb_channels; ch++) { | |
| 226 | ✗ | float *dst = (float *)s->p->extended_data[ch]; | |
| 227 | |||
| 228 | ✗ | for (int i = 0; i < s->kernel_size; i++) | |
| 229 | ✗ | dst[i * s->kernel_size + i] = s->delta; | |
| 230 | } | ||
| 231 | |||
| 232 | ✗ | s->filter_channels = filter_channels_float; | |
| 233 | ✗ | break; | |
| 234 | } | ||
| 235 | |||
| 236 | ✗ | return 0; | |
| 237 | } | ||
| 238 | |||
| 239 | ✗ | static av_cold int init(AVFilterContext *ctx) | |
| 240 | { | ||
| 241 | ✗ | AudioRLSContext *s = ctx->priv; | |
| 242 | |||
| 243 | ✗ | s->fdsp = avpriv_float_dsp_alloc(0); | |
| 244 | ✗ | if (!s->fdsp) | |
| 245 | ✗ | return AVERROR(ENOMEM); | |
| 246 | |||
| 247 | ✗ | return 0; | |
| 248 | } | ||
| 249 | |||
| 250 | ✗ | static av_cold void uninit(AVFilterContext *ctx) | |
| 251 | { | ||
| 252 | ✗ | AudioRLSContext *s = ctx->priv; | |
| 253 | |||
| 254 | ✗ | av_freep(&s->fdsp); | |
| 255 | ✗ | av_frame_free(&s->delay); | |
| 256 | ✗ | av_frame_free(&s->coeffs); | |
| 257 | ✗ | av_frame_free(&s->gains); | |
| 258 | ✗ | av_frame_free(&s->offset); | |
| 259 | ✗ | av_frame_free(&s->p); | |
| 260 | ✗ | av_frame_free(&s->dp); | |
| 261 | ✗ | av_frame_free(&s->u); | |
| 262 | ✗ | av_frame_free(&s->tmp); | |
| 263 | ✗ | } | |
| 264 | |||
| 265 | static const AVFilterPad inputs[] = { | ||
| 266 | { | ||
| 267 | .name = "input", | ||
| 268 | .type = AVMEDIA_TYPE_AUDIO, | ||
| 269 | }, | ||
| 270 | { | ||
| 271 | .name = "desired", | ||
| 272 | .type = AVMEDIA_TYPE_AUDIO, | ||
| 273 | }, | ||
| 274 | }; | ||
| 275 | |||
| 276 | static const AVFilterPad outputs[] = { | ||
| 277 | { | ||
| 278 | .name = "default", | ||
| 279 | .type = AVMEDIA_TYPE_AUDIO, | ||
| 280 | .config_props = config_output, | ||
| 281 | }, | ||
| 282 | }; | ||
| 283 | |||
| 284 | const FFFilter ff_af_arls = { | ||
| 285 | .p.name = "arls", | ||
| 286 | .p.description = NULL_IF_CONFIG_SMALL("Apply Recursive Least Squares algorithm to first audio stream."), | ||
| 287 | .p.priv_class = &arls_class, | ||
| 288 | .p.flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | | ||
| 289 | AVFILTER_FLAG_SLICE_THREADS, | ||
| 290 | .priv_size = sizeof(AudioRLSContext), | ||
| 291 | .init = init, | ||
| 292 | .uninit = uninit, | ||
| 293 | .activate = activate, | ||
| 294 | FILTER_INPUTS(inputs), | ||
| 295 | FILTER_OUTPUTS(outputs), | ||
| 296 | FILTER_QUERY_FUNC2(query_formats), | ||
| 297 | .process_command = ff_filter_process_command, | ||
| 298 | }; | ||
| 299 |